train_results.json

{
    "achieved_tflops_per_gpu": 11.07456719317637,
    "achieved_tflops_per_gpu_theoretical": 670.8051034389872,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.4211060404777527,
    "mfu_percent": 3.549540767043708,
    "mfu_percent_theoretical": 215.00163571762408,
    "total_flos": 1.1298365271930044e+19,
    "train_loss": 0.22232198994150273,
    "train_runtime": 31881.5091,
    "train_samples_per_second": 12.696,
    "train_steps_per_second": 0.198,
    "valid_targets_mean": 15457.9,
    "valid_targets_min": 697
}